Computer-implemented method for determining a temperature of a rotor of an electric machine and for training a hybrid model to determine the temperature

Abstract

The invention relates to a device and a computer-implemented method for determining a temperature of a rotor (104) of an electric machine (101), wherein: a temperature of a stator (102) of the electric machine (101) is measured;, a temperature of the stator (102) in a region of the stator (102) facing the rotor (104) is determined on the basis of the temperature of the stator (102), in particular using a model (114) which is designed to map the temperature of the stator (102) onto the temperature of the stator (102) in the region of the stator (102) facing the rotor (104); a physical model (118) is provided which models a first part of an ordinary differential equation for heat transfer to the rotor (104), in particular from the stator (102), on the basis of the temperature of the stator (102) in the region of the stator (102) facing the rotor (104); a data-driven model (120) is provided which models a second part of the ordinary differential equation on the basis of the temperature of the stator (102) in the region of the stator (102) facing the rotor (104); the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped onto a first part of the heat transfer to the rotor (104) or the temperature of the rotor (104) using the physical model (118); the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped onto a second part of the heat transfer to the rotor (104) or the temperature of the rotor (104) using the data-driven model (120); and the temperature of the rotor (104) is determined on the basis of the first part and the second part, in particular on the basis of a sum of the first part and the second part. The invention also relates to devices and computer-implemented methods for training a hybrid model (116) to determine a temperature of a rotor (104) of an electric machine (101).

Description

[0001] Description

[0002] title

[0003] Computer-implemented method for determining the temperature of a rotor of an electric machine and for training a hybrid model for determining the temperature

[0004] State of the art

[0005] The invention relates to a device and a computer-implemented method for determining a temperature of a rotor of an electrical machine and for training a hybrid model for determining the temperature.

[0006] Disclosure of the invention

[0007] A computer-implemented method for determining a temperature of a rotor of an electrical machine provides that a temperature of a stator of the electrical machine is measured, wherein, depending on the temperature of the stator, a temperature of the stator in a region of the stator facing the rotor is determined, in particular using a model that is designed to map the temperature of the stator to the temperature of the stator in the region of the stator facing the rotor, wherein a physical model is provided that models a first part of an ordinary differential equation for heat transfer to the rotor, in particular from the stator, depending on the temperature of the stator in the region of the stator facing the rotor, wherein a data-driven model is provided that models a second part of the ordinary differential equation depending on the temperature of the stator in the region of the stator facing the rotor,wherein the temperature of the stator in the region of the stator facing the rotor is mapped to a first part of the heat transfer to the rotor using the physical model, wherein the temperature of the stator in the region of the stator facing the rotor is mapped to a second part of the heat transfer to the rotor using the data-driven model, and wherein the temperature of the rotor is determined as a function of the first part and the second part, in particular as a function of a sum of the first part and the second part, or as a function of a variable that characterizes the temperature of the rotor (104), wherein the variable is determined as a function of a sum of the first part and the second part.

[0008] In one example, the physical model is designed to model the heat transfer from the stator and from a coolant for the electric machine to the rotor, wherein the physical model determines the first part of the heat transfer depending on a cooling temperature of the coolant.

[0009] In one example, the data-driven model is designed to model the heat transfer as a function of at least one input variable, in particular the cooling temperature, an input voltage of the electrical machine or a current in the electrical machine, wherein the second part of the heat transfer is determined with the data-driven model as a function of the at least one input variable.

[0010] It can be provided that the hotspot temperature of the stator is determined with a model for determining a hotspot temperature of the stator depending on the temperature of the stator, wherein the model for determining the hotspot temperature of the stator comprises features, and / or wherein the model which is designed to map the temperature of the stator to the temperature of the stator in the region of the stator facing the rotor comprises features, wherein the physical model is designed to determine the first part depending on the features and / or wherein the data-controlled model is designed to determine the second part depending on the features.

[0011] It may be provided that the electric machine is cooled more intensively with the coolant if the size or temperature of the rotor is greater than or equal to a threshold value.It can be provided that for several values ​​of the variable, in particular a 99% quantile of a probability distribution of the variable or for several values ​​of the temperature of the rotor, in particular a 99% quantile of a probability distribution of the temperature of the rotor is determined, wherein a comparison of the quantile of the probability distribution with a threshold value is carried out, and as long as the quantile of the probability distribution is smaller than the threshold value, no increased risk for component protection is to be detected, as long as the quantile of the probability distribution is greater than or equal to the threshold value, an increased risk for component protection is to be detected, and / or the electric machine is to be cooled more intensively with the coolant if the quantile is greater than or equal to the threshold value.

[0012] A computer-implemented method for training a hybrid model for determining a temperature of a rotor of an electrical machine provides that a plurality of tuples are provided, each comprising a temperature of the stator in a region of the stator facing the rotor, and a reference for the temperature of the rotor, wherein a physical model is provided which models a first part of an ordinary differential equation for heat transfer to the rotor, in particular from the stator, as a function of the temperature of the stator in the region of the stator facing the rotor, wherein a data-driven model is provided which models a second part of the ordinary differential equation as a function of the temperature of the stator in the region of the stator facing the rotor, wherein for each tuple,the temperature of the stator in the region of the stator facing the rotor is mapped with the physical model to a first part of the heat transfer to the rotor or the temperature of the rotor, and with the data-driven model to a second part of the heat transfer to the rotor or the temperature of the rotor, and wherein the temperature of the rotor is determined as a function of the first part and the second part, in particular as a function of a sum of the first part and the second part, and wherein the physical model and / or the data-driven model is trained as a function of a deviation of the rotor temperature determined for the respective tuples from the plurality of tuples from the reference for the rotor temperature from the respective tuple. A computer-implemented method for training a hybrid model for determining a temperature of a rotor of an electric machine provides that a plurality of tuples are provided,each comprising a temperature of the stator and a reference for the temperature of the rotor, wherein a physical model is provided which models a first part of an ordinary differential equation for heat transfer to the rotor, in particular from the stator, as a function of a temperature of the stator in a region of the stator facing the rotor, wherein a data-driven model is provided which models a second part of the ordinary differential equation as a function of the temperature of the stator in the region of the stator facing the rotor, wherein for each tuple, the temperature of the stator in the region of the stator facing the rotor is determined as a function of the temperature of the stator, in particular with a model which is designed to map the temperature of the stator to the temperature of the stator in the region of the stator facing the rotor,wherein the temperature of the stator in the region of the stator facing the rotor is mapped to a first part of the heat transfer to the rotor or the temperature of the rotor using the physical model, and to a second part of the heat transfer to the rotor or the temperature of the rotor using the data-driven model, and wherein the temperature of the rotor is determined as a function of the first part and the second part, in particular as a function of a sum of the first part and the second part, or as a function of a variable that characterizes the temperature of the rotor, the variable being determined as a function of a sum of the first part and the second part, and wherein the physical model and / or the data-driven model is trained as a function of a deviation of the temperature of the rotor determined for the respective tuples from the plurality of tuples from the reference for the temperature of the rotor from the respective tuple.

[0013] The training method can provide that the physical model is designed to model the heat transfer from the stator and from a coolant for the electric machine to the rotor, wherein the first part of the heat transfer is determined with the physical model as a function of a cooling temperature of the coolant, and wherein the tuples each comprise the cooling temperature of the coolant. The training method can provide that the data-driven model is designed to model the heat transfer as a function of at least one input variable, in particular the cooling temperature, an input voltage of the electric machine, or a current in the electric machine, wherein the second part of the heat transfer is determined with the data-driven model as a function of the at least one input variable, and wherein the tuples each comprise the at least one input variable.

[0014] The training method can provide that the hotspot temperature of the stator is determined using a model for determining a hotspot temperature of the stator depending on the temperature of the stator, wherein the model for determining the hotspot temperature of the stator comprises features, and / or wherein the model which is designed to map the temperature of the stator to the temperature of the stator in the region of the stator facing the rotor comprises features, wherein the physical model is designed to determine the first part depending on the features and / or wherein the data-controlled model is designed to determine the second part depending on the features.

[0015] The method for training can provide that the hybrid model for a device for determining a temperature of a rotor of an electric machine comprises parameters, wherein the hybrid model is trained in a training with a computing environment provided outside the device, wherein updated parameters are determined in the training, wherein the device is connected to the computing environment outside the device for updating the parameters via a communication connection, in particular a wireless communication connection, wherein the parameters in the device are replaced by the updated parameters.

[0016] A device for determining a temperature of a rotor of an electric machine or for training a hybrid model for determining a temperature of a rotor of an electric machine is configured to carry out the method for determining the temperature or for training. In one example, the device comprises at least one processor and at least one non-volatile memory, wherein the at least one processor is configured to execute instructions, upon the execution of which by the at least one processor, the device executes the method, and wherein at least one non-volatile memory stores the instructions.

[0017] A computer program comprising computer-executable instructions, the execution of which by the computer causes the method to run on the computer, may be provided.

[0018] Further advantageous embodiments can be found in the following description and the drawing. The drawing shows:

[0019] Fig. 1 is a first schematic representation of a device, Fig. 2 is a second schematic representation of the device, Fig. 3 is a first flowchart with steps of a method for determining a temperature of a rotor of an electrical machine, Fig. 4 is a second flowchart with steps of a first method for training a hybrid model for determining the temperature, Fig. 5 is a third flowchart with steps of a second method for training the hybrid model for determining the temperature.

[0020] Figure 1 shows a device 100 for determining a quantity x T an electrical machine 101 with a stator 102 and a rotor 104, which has a temperature T R of the rotor 104, shown schematically. The size x TIn an example, the temperature T R of the rotor 104. In an example, the size x T a state which the temperature T R of the rotor 104. In one example, the size x T a state which the temperature T R of the rotor 104. It can be provided that the size x T is multidimensional and includes operating states of the electrical machine 101. It can be provided that the size x T the temperature T R of the rotor 104. It can be provided that the temperature T R of the rotor 104 from the size x T is calculable. In the example, the device 100 comprises an input for a measured temperature T s of the stator 102. In the example, a temperature sensor 106 is provided, which is designed to measure the temperature T s of the stator 102.

[0021] The device 100 optionally comprises a first device 108 for detecting an anomaly or for checking plausibility.

[0022] The first device 108 is designed to detect the anomaly of the temperature T s to recognize or the plausibility of the temperature T s to check.

[0023] The first device 108 is designed to measure the temperature T s of the stator 102, provided no anomaly is detected and / or the temperature T s of the stator 102 is recognized as plausible in the plausibility check, and the temperature T s of the stator 102 otherwise not to be provided.

[0024] The device 100 optionally comprises a second device 118 for detecting an anomaly or for checking plausibility.

[0025] The second device 118 is designed to detect the anomaly of input variables , ...,I nto recognize or to check the plausibility of input variables / 1; ...,I n to check.

[0026] The second device 118 is designed to convert the input variables / 1; ...,I n provided no anomaly is detected and / or the input variables / 1; ...,I n are recognized as plausible in the plausibility check, and the input variables / 1; ...,I n otherwise not to be provided.

[0027] The input variables / 1; ...,I n are, for example, an input voltage, a cooling temperature T K , a current of the electrical machine 101 .

[0028] The device 100 comprises a first model 112. The first model 112 is for determining a hotspot temperature T H of the stator 102 depending on the temperature T measured by the temperature sensor 106 sof the stator 102. The first model 112 can be an empirical, physical, or data-driven model. The first model 112 can be used to extract features F, such as model states, and provide them.

[0029] The device 100 comprises a second model 114. The second model 114 is for determining a temperature T SR of the stator 102 at a region of the stator 102 facing the rotor 104 depending on the measured temperature T s of the stator 102 and the input variables / 1; ...,I n The second model 114 can be an empirical, physical, or data-driven model. The second model 114 can be used to extract features F, such as model states, and provide them.

[0030] The first model 112 may have parameters that depend on pairs of temperatures T measured by the temperature sensor 106. s and one of the measured temperatures T s associated measured hotspot temperature of the stator 102, are or will be learned or validated.

[0031] The second model 114 may have parameters that depend on pairs of measured temperatures of the T s and one of the measured temperatures T s associated measured temperature of the stator 102 at the region of the stator 102 facing the rotor 104, are or will be learned or validated.

[0032] The model 112 for determining the hotspot temperature T H and / or the model 114 for determining the temperature T SR includes features F.

[0033] The features F can be states of the model 112 to determine the hotspot temperature T Hor model 114 to determine the temperature T SR or control variables of Model 112 to determine the hotspot temperature T H or model 114 to determine the temperature T SR be.

[0034] For example, the first model 112 comprises a first artificial neural network, wherein the features F are provided to a layer of the first artificial neural network at a time t. For example, the second model 114 comprises a second artificial neural network, wherein the features F are provided to a layer of the second artificial neural network at a time t.

[0035] The apparatus 100 includes a hybrid model 116 for a hybrid ordinary differential equation.

[0036] The hybrid ordinary differential equation is e.g. x T = (a, x T , F, t) + g(ß, x T, F, t) with a physical part f and a data-driven part g, where the size x T is a state that characterizes the temperature of the rotor 104 implicitly or explicitly, particularly in Kelvin. In the example, a change in the quantity x characterizes T a heat transfer at the rotor 104.

[0037] The hybrid ordinary differential equation is defined depending on the characteristics F from the first model 112 and / or from the second model 114.

[0038] The physical part f is modeled, for example, by a physical model 118 with parameters a of the physical part f, where the physical part f has the size x T at a time t depending on the parameters a of the physical part f to an output of the physical part f.

[0039] The output of the physical part f in the example depends on the cooling temperature T K from the input variables ß, For example, a heat flow from a coolant with the cooling temperature T K to the rotor 104 depending on a difference between cooling temperature T K and the last calculated temperature of the rotor 104. In the example where the size x T the temperature of the rotor 104, the difference T K — x T In the example where the size x T characterizes the temperature of the rotor 104, the difference T K — T R determined, where T R the temperature of the rotor 104, which depends on the size x T The heat flow in the example is determined by multiplying the difference between the cooling temperature T Kand the last calculated temperature of the rotor 104 with a specific thermal conductivity of the coolant. The specific thermal conductivity of the coolant is calculated in the example within the physical part f based on the parameter a and the characteristics F.

[0040] Characteristics can include, for example, the flow rate Q K of the coolant and the temperature of the coolant T K according to the calculation rule

[0041] G K = («1 + «2 QK 3 + a A) 1 serve.

[0042] It may be provided that the size x T is described probabilistically by taking uncertainties of the temperature T R of the rotor 104 using a Laplace approximation of the parameters a, where the size x T depending on the Laplace approximation of the parameter a. This means that the size x Tand the uncertainties for the size x T certainly.

[0043] For example, a heat flow from the stator 102 to the rotor 104 is determined depending on a difference between the last calculated temperature T s of the stator 102 and the last calculated temperature of the rotor 104. In the example where the size x T the temperature of the rotor 104, the difference T s — x T In the example where the size x T characterizes the temperature of the rotor 104, the difference T s — T R determined, where T R the temperature of the rotor 104, which depends on the size x T The heat flow in the example is determined by multiplying the difference between the last calculated temperature T sof the stator 102 and the last calculated temperature of the rotor 104 with a specific thermal conductivity of the stator 102. The specific thermal conductivity of the stator 102 is determined in the example for the stator 102 using the physical parameters a and the characteristics F by the rule

[0044] G s = (a5n“ 6 ) -1 calculated. Here, n is the speed of the electric machine.

[0045] Furthermore, in the example, in the context of the physical part f, the power loss P L of the rotor 104 is calculated. This is done based on the characteristics F and the physical parameters a. For example, the power loss is calculated using a characteristic map whose parameters are contained in the physical parameters a. In the example, the input variables for the characteristic map are the phase currents I d and I q of the electrical machine and the speed n of the electrical machine.

[0046] The temperature differential of the temperature of the rotor 104 calculated by the physical part f in the example is therefore f(a,x T ,F, t) = P L (a,F) + G s (a,n(T s - T R ) + G K a, F^T K - T R ) where G s the transfer function for the heat flow from the stator 102 to the rotor 104 and G K the transfer function for the heat flow from the coolant to the rotor 104.

[0047] The described transfer functions G s and G K represent an illustrative example of equations in the physical part f. The concept is also applicable to other equations for these quantities.

[0048] The data-driven part g is modeled, for example, by a data-driven model 120 with parameters ß of the data-driven part g, where the data-driven part g has the size x Tat a time t, depending on the parameters ß of the data-driven part g, to an output of the data-driven part g. This output serves to correct the predictions of the physical part f by data-drivenly mapping effects not represented in the physical part f.

[0049] In the example, the expenditures are estimated to change x T added. Another way of combining the outputs, e.g., a weighted addition, is also possible.

[0050] Predictions of the hybrid model 116 for the course of the size x T are determined, for example, using numerical methods for solving ordinary differential equations. These are methods for the temporal discretization of ordinary differential equations, such as the Euler method.

[0051] The hybrid model 116 is designed, the size x Twhich solves the differential equation at time t, depending on the measured temperature T s , which for this particular temperature T SR , the characteristics F, and the input variables / 1; to be provided at time t.

[0052] The parameters a,ß are or will be learned or validated depending on tuples, which are measured input variables ß, a measured temperature T R of the rotor 104, a measured temperature T s of the stator 102, a measured hotspot temperature of the stator 102, a temperature of the stator 102 at a region of the stator 102 facing the rotor 104, which are associated with one another in particular at time t.

[0053] For example, the parameters a are determined using a least squares method. For example, the parameters ß are determined using a gradient descent method. In the example where the variable x T the temperature of the rotor 104, for example, the parameters a,ß are determined for which a deviation of the size x T of the measured temperature of the rotor 104 for the tuples is as small as possible.

[0054] In the example where the size x T characterizes the temperature of the rotor 104, the procedure is accordingly, whereby the parameters a,ß are determined for which a deviation of the value for the respective size x T certain temperature T R of the measured temperature of the rotor 104 for the tuples is as small as possible.

[0055] In the example, the device 100 is designed to determine the temperature of the rotor 104 of the electric machine 101 or to determine the temperature for training the hybrid model 116 for determining the temperature of the rotor 104 of the electric machine 101 in a method for determining the temperature or for training. Training can be provided in a vehicle and retraining, specifically in a computing environment provided outside the vehicle, i.e., outside the device 100, e.g., a cloud computing environment. For example, the device 100 is set up to update the parameters a and / or the parameters ß in the device 100 via a wireless communication connection, i.e., "over the air." For example, the tuples are transmitted from the vehicle to the computing environment outside the device 100, and the parameters a and / or the parameters ß, i.e.,New information is learned centrally in the computing environment outside the device 100. For example, the device 100 is provided in various vehicles, wherein the computing environment is configured to learn the parameters a and / or parameters ß depending on the tuples from the various vehicles and to make the learned parameters a and / or parameters ß available to the vehicles, ie, the devices 100 in the respective vehicles. This enables a more precise determination of the variable x. T enabled.

[0056] The device 100 is designed, for example, on the basis of this more precise determination of the state of the size x T to operate at least one component or system in the vehicle in an optimized manner. In one example, the component or system includes the rotor 104. In one example, the temperature T Rof the rotor 104, the temperature of the component or system. For example, the device 100 is designed to be based on a quantified probability p (e.g., p < 0.0001%) or quantiles of a probability distribution of values ​​of size x T or the temperature T R of the rotor 104 to optimize an operating strategy for operating the component or system, e.g. to avoid overwriting critical temperatures and to avoid damage to the component or system.

[0057] In one example, the device 100 is configured to cool the electric machine more strongly with the coolant when the size x T or the temperature T R of the rotor 104 is greater than a threshold value T TH is.

[0058] For example, the device 100 is designed to be able to calculate for several values ​​of the quantity x T a 99% quantile T q the probability distribution of size xT , ie a confidence interval for the size x T , to determine.

[0059] For example, the device 100 is designed to be able to measure several values ​​of the temperature T R of the rotor 104, for example, 99% quantile T q the probability distribution of the temperature T R of the rotor 104, ie a confidence interval for the temperature T R of the rotor 104.

[0060] For example, the plurality of values ​​are each determined for different points in time depending on the temperatures of the stator 102 determined at the different points in time. For example, the plurality of values ​​are each determined for the different points in time depending on the input variables / 1; ...,I n For example, the device 100 is designed to perform an adjustment of the quantile T q the probability distribution with the threshold TTH For example, the device 100 is designed to perform as long as the quantile T q the probability distribution is smaller than the threshold T TH is to detect no increased risk for component protection. For example, the device 100 is designed as long as the quantile T q the probability distribution is greater than or equal to the threshold T TH is to detect an increased risk for component protection. For example, the device 100 is designed to cool more intensively with the coolant when the quantile T q the probability distribution is greater than or equal to the threshold T TH is.

[0061] In Figure 2, a part of the device 100 is shown schematically.

[0062] The device 100 includes at least one processor 202 and at least one non-volatile memory 204.

[0063] The at least one processor 202 is configured to execute instructions, upon execution of which by the at least one processor 202 the device 100 executes the respective method.

[0064] The at least one non-volatile memory 204 stores the instructions.

[0065] It can be provided that the device 100 has an interface 206 to the temperature sensor 106 and / or for the input variables / 1; ...,I n It can be provided that the device 100 includes the temperature sensor 106.

[0066] Figure 3 shows a flow chart with steps of the particular computer-implemented method for determining the temperature T R of the rotor 104.

[0067] The method for determining the temperature T Rof the rotor 104 includes a step 300. In step 300, the physical model 118 and the data-driven model 120 are provided.

[0068] The method for determining the temperature T R of the rotor 104 includes a step 302.

[0069] In step 302, the temperature T s of the stator 102.

[0070] The method for determining the temperature T R of the rotor 104 includes a step 304.

[0071] In step 304, depending on the temperature T s of the stator the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104.

[0072] The temperature T SR of the stator 102 in the area of ​​the stator 102 facing the rotor 104 is determined in the example with the second model 114.

[0073] It can be provided that with the first model 112 the hotspot temperature T Hof the stator 102 is determined.

[0074] The method for determining the temperature T R of the rotor 104 includes a step 306.

[0075] In step 306, the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104 is mapped to a first part of the heat transfer to the rotor 104 using the physical model 118. It can be provided that the first part of the heat transfer is mapped to the physical model 118 depending on the cooling temperature T K of the coolant.

[0076] In step 306, the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104 is mapped to a second part of the heat transfer to the rotor 104 using the data-driven model 120. It can be provided that the second part of the heat transfer is mapped to the rotor 104 using the data-driven model 120 depending on the at least one input variable / 1; ...,In is determined.

[0077] In the example, the first part and the second part are parts of size x T .

[0078] It can be provided that the first model 112 and / or the second model 114 comprises features F.

[0079] It can be provided that the first part is determined with the physical model 118 depending on the features F.

[0080] It may be provided that the second part is determined with the data-driven model 120 depending on the features F.

[0081] The method for determining the temperature T R of the rotor 104 includes a step 308.

[0082] In step 308, the temperature T R of the rotor 104 depending on the first part and the second part.

[0083] For example, the temperature T R of the rotor 104 is determined depending on a sum of the first part and the second part.

[0084] In an example, the size x T the temperature T R of the rotor 104, ie the temperature T R of the rotor 104 is determined depending on the solution of the ordinary differential equation depending on the sum of the first part and the second part.

[0085] In an example, the quantity x characterizes T the temperature T R of the rotor 104, ie the temperature T R of the rotor 104 depends on the size x T which is the solution of the ordinary differential equation dependent on the sum of the first part and the second part.

[0086] The method may provide that the electric machine is cooled more strongly with the coolant if the size x T or the temperature T R of the rotor 104 is greater than a threshold value T TH The method can provide that for several values ​​of the quantity x Ta 99% quantile T q a probability distribution, e.g. a normal distribution or a Student distribution, of size x T The method may provide that for several values ​​of temperature T R of the rotor 104, for example, 99% quantile T q a probability distribution, e.g. a normal distribution or a Student distribution, the temperature T R of the rotor 104. The quantile T q represents a confidence interval.

[0087] The method may include an adjustment of the quantile T q the probability distribution with the threshold T TH to execute.

[0088] The method may provide that as long as the quantile T q the probability distribution is smaller than the threshold T TH no increased risk to component protection is detected.

[0089] The method may provide that as long as the quantile T q the probability distribution is greater than or equal to the threshold T TH an increased risk to component protection is detected.

[0090] The method may provide for more intense cooling with the coolant if the quantile T q the probability distribution is greater than or equal to the threshold T TH is.

[0091] Figure 4 shows a flowchart with steps of a particularly computer-implemented first method for training the hybrid model 116 for determining the temperature of the rotor 104.

[0092] The first method for training the hybrid model 116 includes a step 400.

[0093] In step 400, a plurality of tuples are provided, each containing a temperature T SRof the stator 102 in a region of the stator 102 facing the rotor 104, and a reference for the temperature of the rotor 104. The first method for training the hybrid model 116 includes a step 402.

[0094] In step 402, the physical model 118 and the data-driven model 120 are provided

[0095] The first method for training the hybrid model 116 includes a step 404.

[0096] In step 404, for each tuple, the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104 is mapped with the physical model 118 to a first part of the heat transfer to the rotor 104 or the temperature of the rotor 104.

[0097] In step 404, for each tuple, the temperature T SRof the stator 102 in the region of the stator 102 facing the rotor 104 is mapped to a second part of the heat transfer to the rotor 104 or the temperature of the rotor 104 using the data-driven model 120.

[0098] The first method for training the hybrid model 116 includes a step 406.

[0099] In step 406, the temperature T R of the rotor 104 depending on the first part and second part determined for the respective tuple.

[0100] In the example, the temperature T R of the rotor 104 is determined depending on a sum of the respective first part and the second part.

[0101] The first method for training the hybrid model 116 includes a step 408.

[0102] In step 408, and wherein the physical model 118 and / or the data-driven model 120 is / are dependent on a deviation of the temperature T determined for the respective tuples from the plurality of tuples R of the rotor 104 is trained from the reference for the temperature of the rotor 104 from the respective tuple. Figure 5 shows a flowchart with steps of a particularly computer-implemented second method for training the hybrid model 116 for determining the temperature T R of the rotor 104.

[0103] The second method for training the hybrid model 116 includes a step 500.

[0104] In step 500, a plurality of tuples are provided, each containing a temperature T s of the stator 102 and a reference for the temperature of the rotor 104.

[0105] The second method for training the hybrid model 116 includes a step 502.

[0106] In step 502, the physical model 118 and the data-driven model 120 are provided.

[0107] The second method for training the hybrid model 116 includes a step 504.

[0108] In step 504, for each tuple, the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104 depending on the temperature T s of the stator 102. The temperature T SR of the stator 102 in the area of ​​the stator 102 facing the rotor 104 is determined in the example with the second model 114.

[0109] The second method for training the hybrid model 116 includes a step 506.

[0110] In step 506, the temperature T SRof the stator 102 in the region of the stator 102 facing the rotor 104 is mapped with the physical model 118 to a first part of the heat transfer to the rotor 104 or the temperature of the rotor 104. In step 506, the temperature T SR of the stator 102 in the region of the stator 102 facing the rotor 104 is mapped to a second part of the heat transfer to the rotor 104 or the temperature of the rotor 104 using the data-driven model 120.

[0111] The second method for training the hybrid model 116 includes a step 508.

[0112] In step 508, the temperature T R of the rotor 104 depending on the first part and the second part.

[0113] The temperature T R of the rotor 104 is determined in the example depending on a sum of the first part and the second part.

[0114] The second method for training the hybrid model 116 includes a step 510.

[0115] In step 510, the physical model 118 and / or the data-driven model 120 is / are modified depending on a deviation of the temperature T determined for the respective tuples from the plurality of tuples. R of the rotor 104 from the reference for the temperature of the rotor 104 from the respective tuple.

[0116] In training, it can be provided that with the physical model 118 the first part of the heat transfer is dependent on a cooling temperature T K of the coolant is determined. In training, the tuples in this case include, for example, the cooling temperature T K of the coolant

[0117] In the training, it can be provided that the data-driven model 120 is used to calculate the second part of the heat transfer depending on the at least one input variable / 1; ...,I nIn training, the tuples in this case each contain at least one input variable / 1;

[0118] During training, it may be provided that the hotspot temperature of the stator 102 is determined using the first model 112. During training, it may be provided that the first model 112 and / or the second model 114 includes features F

[0119] During training, it may be provided that the physical model 118 determines the first part depending on the features F and / or the data-driven model 120 determines the second part depending on the features F.

[0120] For example, during training, the parameters a are determined using the least squares method. For example, during training, the parameters ß are determined using the gradient descent method.

[0121] The method for training may provide that the hybrid model 116 is trained in training with a computing environment provided external to the device 100.

[0122] During training, for example, updated parameters a and / or parameters ß are determined.

[0123] For example, the device 100 is connected to the computing environment outside the device 100 via a wireless communication connection for updating the parameters. The parameters a and / or parameters ß in the device are replaced by the updated parameters, for example, via the communication connection.

Claims

Claims 1. A computer-implemented method for determining a temperature of a rotor (104) of an electrical machine (101), characterized in that a temperature of a stator (102) of the electrical machine (101) is measured (302), wherein, depending on the temperature of the stator (102), a temperature of the stator (102) in a region of the stator (102) facing the rotor (104) is determined (304), in particular using a model (114) that is designed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein a physical model (118) is provided (300) that defines a first part of an ordinary differential equation for heat transfer to the rotor (104), in particular from the stator (102), depending on the temperature of the stator (102) in the region of the stator facing the rotor (104). (102) modeled, whereby a data-driven model (120) is provided (300),which models a second part of the ordinary differential equation as a function of the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (306) with the physical model (118) to a first part of the heat transfer to the rotor (104), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (306) with the data-driven model (120) to a second part of the heat transfer to the rotor (104), and wherein the temperature of the rotor (104) is determined (308) as a function of the first part and the second part, in particular as a function of a sum of the first part and the second part, or as a function of a variable that characterizes the temperature of the rotor (104), wherein the size is determined depending on a sum of the first part and the second part., 2. Method according to claim 1, characterized in that the physical model (118) is designed to model the heat transfer from the stator (102) and from a coolant for the electrical machine (101) to the rotor (104), wherein the first part of the heat transfer is determined with the physical model (118) as a function of a cooling temperature of the coolant (306).

3. Method according to one of the preceding claims, characterized in that the data-controlled model (120) is designed to calculate the heat transfer as a function of at least one input variable ( / 1; ..., / „), in particular the cooling temperature, an input voltage of the electrical machine (101) or a current in the electrical machine (101), wherein the data-controlled model (120) is used to determine the second part of the heat transfer as a function of the at least one input variable (306).

4. Method according to one of the preceding claims, characterized in that the hotspot temperature of the stator (102) is determined (304) with a model (112) for determining a hotspot temperature of the stator (102) depending on the temperature of the stator (102), wherein the model (112) for determining the hotspot temperature of the stator (102) comprises features (F), and / or wherein the model (114) which is designed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) comprises features, wherein the physical model (118) is designed to determine the first part depending on the features and / or wherein the data-controlled model (120) is designed to determine the second part depending on the features.

5. Method according to one of the preceding claims, characterized in that the electric machine is cooled more intensively with the coolant when the size or the temperature of the rotor (104) is greater than or equal to a threshold value.

6. Method according to one of the preceding claims, characterized in that for several values ​​of the size, in particular a 99% quantile of a probability distribution of the size or for several values ​​of the Temperature of the rotor (104) a particularly 99% quantile of a probability distribution of the temperature of the rotor (104) is determined, wherein a comparison of the quantile of the probability distribution with a threshold value is carried out, and as long as the quantile of the probability distribution is smaller than the threshold value, no increased risk for component protection is to be detected, as long as the quantile of the probability distribution is greater than or equal to the threshold value, an increased risk for component protection is to be detected, and / or the electric machine (101) is to be cooled more intensively with the coolant if the quantile of the probability distribution is greater than or equal to the threshold value.

7. A computer-implemented method for training a hybrid model (116) for determining a temperature of a rotor (104) of an electrical machine (101), characterized in that a plurality of tuples are provided (400), each comprising a temperature of the stator (102) in a region of the stator (102) facing the rotor (104), and a reference for the temperature of the rotor (104), wherein a physical model (118) is provided (402) which models a first part of an ordinary differential equation for heat transfer to the rotor (104), in particular from the stator (102), as a function of the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein a data-driven model (120) is provided (402) which models a second part of the ordinary differential equation as a function of the temperature of the stator (102) in the rotor (104) facing area of ​​the stator (102), where for each tuple,the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (404) with the physical model (118) to a first part of the heat transfer to the rotor (104) or the temperature of the rotor (104) and with the data-controlled model (120) to a second part of the heat transfer to the rotor (104) or the temperature of the rotor (104), and wherein the temperature of the rotor (104) is determined (406) as a function of the first part and the second part, in particular as a function of a sum of the first part and the second part, and wherein the physical model (118) and / or the data-controlled model (120) as a function of a deviation of the temperature determined for the respective tuples from the plurality of tuples, Rotor (104) is trained (408) by the reference for the temperature of the rotor (104) from the respective tuple.

8. A computer-implemented method for training a hybrid model (116) for determining a temperature of a rotor (104) of an electrical machine (101), characterized in that a plurality of tuples are provided (500), each comprising a temperature of the stator (102) and a reference for the temperature of the rotor (104), wherein a physical model (118) is provided (502) which models a first part of an ordinary differential equation for heat transfer to the rotor (104), in particular from the stator (102), as a function of a temperature of the stator (102) in a region of the stator (102) facing the rotor (104), wherein a data-driven model (120) is provided (502) which models a second part of the ordinary differential equation as a function of the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein for each tuple,the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is determined (504) as a function of the temperature of the stator (102), in particular with a model (114) which is designed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104), wherein the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) is mapped (506) with the physical model (118) to a first part of the heat transfer to the rotor (104) or the temperature of the rotor (104) and with the data-controlled model (120) to a second part of the heat transfer to the rotor (104) or the temperature of the rotor (104), and wherein the temperature of the rotor (104) is determined as a function of the first part and the second part is determined (508), in particular depending on a sum of the first part and the second part, or depending on a size,which characterizes the temperature of the rotor (104), wherein the quantity is determined as a function of a sum of the first part and the second part, and wherein the physical model (118) and / or the data-driven model (120) is determined as a function of a deviation of the temperature of the rotor determined for the respective tuples from the plurality of tuples, Rotor (104) is trained (510) by the reference for the temperature of the rotor (104) from the respective tuple.

9. The method according to one of claims 7 or 8, characterized in that the physical model (118) is designed to model the heat transfer from the stator (102) and from a coolant for the electrical machine (101) to the rotor (104), wherein the physical model (118) determines the first part of the heat transfer as a function of a cooling temperature of the coolant (404; 506), and wherein the tuples each comprise the cooling temperature of the coolant.

10. The method according to one of claims 7 to 9, characterized in that the data-driven model (120) is designed to calculate the heat transfer as a function of at least one input variable ( / 1; ... , / „), in particular the cooling temperature, an input voltage of the electrical machine (101) or a current in the electrical machine (101), wherein the data-controlled model (120) determines the second part of the heat transfer as a function of the at least one input variable ( , ... , / „) is determined (404; 506), and wherein the tuples each contain at least one input variable ( / 1; ... , / „).

11. Method according to one of claims 7 to 10, characterized in that the hotspot temperature of the stator (102) is determined (204) with a model (112) for determining a hotspot temperature of the stator (102) depending on the temperature of the stator (102), wherein the model (112) for determining the hotspot temperature of the stator (102) comprises features, and / or wherein the model (114) which is designed to map the temperature of the stator (102) to the temperature of the stator (102) in the region of the stator (102) facing the rotor (104) comprises features, wherein the physical model (118) is designed to determine the first part depending on the features and / or wherein the data-controlled model (120) is designed to determine the second part depending on the features.

12. Method according to one of claims 7 to 11, characterized in that the hybrid model (116) for a device (100) for determining a temperature of a rotor (104) of an electrical machine (101) Parameters, wherein the hybrid model (116) is trained in a training with a computing environment provided outside the device (100), wherein updated parameters are determined in the training, wherein the device (100) is connected to the computing environment outside the device (100) via a communication link, in particular a wireless one, for updating the parameters, wherein the parameters in the device are replaced by the updated parameters.

13. Device (100) for determining a temperature of a rotor (104) of an electrical machine (101) or for training a hybrid model (116) for determining a temperature of a rotor (104) of an electrical machine (101), characterized in that the device (100) is designed to carry out the method according to one of claims 1 to 12.

14. Device (100) according to claim 13, characterized in that the device (100) comprises at least one processor (202) and at least one non-volatile memory (204), wherein the at least one processor (302) is designed to execute instructions, upon execution of which by the at least one processor (202), the device (100) carries out the method according to one of claims 1 to 9, and wherein at least one non-volatile memory (204) stores the instructions.

15. A computer program, characterized in that the computer program comprises computer-executable instructions, the execution of which by the computer causes the method according to one of claims 1 to 12 to run on the computer.

Images